Behavioural Processes 95 (2013) 50–59

Contents lists available at SciVerse ScienceDirect

Behavioural Processes
journal homepage: www.elsevier.com/locate/behavproc

Effects of motion on time perception夽

Andréia Kroger-Costa a,∗ , Armando Machado a , Jorge A. Santos b
a
School of Psychology, University of Minho, Animal Learning and Behavior Laboratory, Braga, Portugal
b
School of Psychology and Centro Algoritmi, School of Engineering, University of Minho, Centro de Computação Gráfica, Portugal

a r t i c l e i n f o a b s t r a c t

Article history: To investigate the effect of motion on time perception, participants were asked to perform either a tem-
Received 11 October 2012 poral discrimination task or a temporal generalization task while running or standing still on a treadmill.
Received in revised form 4 February 2013 In the temporal discrimination (bisection) task, 10 participants were exposed to two anchor stimuli, a
Accepted 5 February 2013
300-ms Short tone and a 700-ms Long tone, and then classified intermediate durations in terms of their
similarity to the anchors. In the temporal generalization task, 10 other participants were exposed to a
Keywords:
standard duration (500 ms) and then judged whether or not a series of comparison-durations, ranging
Active motion
from 300 ms to 700 ms, had the same duration as the standard. The results showed that in the temporal
Attentional Allocation Model
Dual-task paradigm
bisection task the participants produced more “Long” responses under the dual-task condition (temporal
Temporal bisection judgments + running) than under the single-task condition (temporal judgments only). In the temporal
Temporal generalization generalization task, accuracy in the temporal judgments was lower in the dual-task condition than the
Time perception single-task condition. These results are discussed in the light of dual-task paradigm and of the Scalar
Scalar Expectancy Theory Expectancy Theory (SET).
This article is part of a Special Issue entitled: SQAB 2012.
© 2013 Elsevier B.V. All rights reserved.

1. Introduction finally, in discrimination tasks (the most common are temporal

To behave adaptively, animals and humans need to be sensitive
to the temporal properties of the environment and the tempo-
ral properties of their behavior (Grondin, 2010; Michon, 1993).
The study of the ability to discriminate the order and duration of
events (time perception), to differentiate the temporal properties of
actions (temporal differentiation), and, more generally, to regulate
behavior on the basis of the temporal attributes of the environment
is called interval timing, or simply timing (see Richelle and Lejeune,
1980; Wearden, 2008).
Timing has been usually studied by means of four basic tasks:
(1) verbal estimation; (2) production; (3) reproduction and (4) dis-
crimination. In verbal estimation tasks, the participant is asked to
estimate verbally the duration of a standard stimulus. In produc-
tion tasks, the participant is asked to produce a standard duration
specified verbally by the experimenter (e.g., “Press a button for
10 seconds”). In reproduction tasks, the participant is asked to
reproduce a standard duration presented by experimenter. And
夽 Portions of the data were presented at 22nd Congress of Spanish Society for
Comparative Psychology (Almería, Spain, 2010), and the TIMELY Workshop on
the “Psychophysical, Computational, and Neuroscience Models of Time Perception”
(Groningen, Netherlands, 2011).
∗ Corresponding author.
E-mail addresses: andreiakc@gmail.com (A. Kroger-Costa),
armandom@psi.uminho.pt (A. Machado), jorge.a.santos@psi.uminho.pt
(J.A. Santos).
bisection and temporal generalization), a stimulus is presented by
the experimenter and the participant is asked to classify it as more
similar to a Short or to a Long standard (temporal bisection), or
to classify it as Equal to or Different from a standard (temporal
generalization). We refer to the subject’s response generically as
“judgment” and the duration the subject is asked to judge generi-
cally as “standard” (Bindra and Wakesberg, 1956).
To account for the results obtained in timing tasks, researchers
have use a variety of concepts (e.g., under and overestimation,
subjective and objective time units, speed of internal and external
clocks, etc.), but these concepts have not always been used in clear
and consistent ways (see Bindra and Wakesberg, 1956). To avoid
confusion, henceforth we adopt the conceptual framework of the
Scalar Expectancy Theory (SET). SET is an information-processing
model that postulates an internal clock with three components:
a pacemaker–accumulator unit, a memory store, and a compara-
tor/decision component. At the onset of the standard stimulus, the
pacemaker starts to generate pulses at a high rate, which pulses
are then added in the accumulator. When the standard ends, the
number of pulses in the accumulator is saved in a memory store.
Because the system is noisy (e.g., the speed of the pacemaker
varies across trials), the number in the accumulator at the end of
the same stimulus will vary across trials. Hence, the memory for
the stimulus duration is represented not by a number but by a
distribution of different numbers – typically a normal distribution.
To make a temporal judgment, the subject compares the number
currently in the accumulator (a measure of elapsed time) with

0376-6357/$ – see front matter © 2013 Elsevier B.V. All rights reserved.
http://dx.doi.org/10.1016/j.beproc.2013.02.002
 A. Kroger-Costa et al. / Behavioural Processes 95 (2013) 50–59
a number sampled from its memory store (a measure of the
standard’s duration). The judgment will depend on how close the
two numbers are (see Gibbon, 1977, 1991; Grondin, 2010; Lejeune
and Wearden, 2006; Wearden, 1991b).
Some results from timing tasks are easily accommodated within
SET. We provide three examples. If a participant produces a 20-s
interval when asked to produce a 15-s interval (the standard), one
can conclude that the number of pulses the subject associates with
1 (subjective) second require more than 1 second to be generated,
or that the subjective time unit is longer than the physical time unit.
One could also say that the internal clock is slower than the external
clock, or that the subject overestimates the label “15 seconds”, but
underestimates the physical interval of 20 s.
As a second example, consider a subject that sees an experi-
menter produce a 15-s standard and then, when reproducing it,
generates a 20-s interval. In this case, we cannot conclude any-
thing about the relative speeds of the internal and the external
clocks or about the relative lengths of the psychological and the
physical time units, for differences in speeds or units can still yield
accurate reproductions of the standard. However, we can conclude
that the speed of the internal clock decreased from the moment
the standard was presented to the moment it was reproduced. To
see this, suppose that initially the pacemaker emitted 1 pulse per
second on the average. Then, at the end of the standard, the accumu-
lator would register 15 pulses and this number would be saved in
memory. If subsequently the speed of the pacemaker decreased to
0.75 pulses per second, the subject would need 20 s on the average
to reproduce the 15 pulses associated with the standard.
Finally, in a discrimination task, suppose that the subject learns
to classify a 0.5-s standard as “Short” and a 1.0-s standard as “Long”
on the basis of the different number of pulses associated with each
stimulus, say, 5 and 10 on the average. Subsequently, if the speed
of the internal clock increases, the standard previously classified
as “Short” may now be classified as “Long” because the number
of pulses emitted in its presence may have increased from 5 to
10. We could say that the standards were now overestimated. SET
then provides a clear conceptual framework to describe and analyze
timing data.
In this article we use SET to understand how timing may be
affected by motion. Moving stimuli tend to be perceived as longer
than static stimuli (see, Kanai et al., 2006; Matthews, 2011) and
similar effects occur when it is the participant, instead of the stim-
ulus, that moves (Binetti et al., 2010; Capelli et al., 2007; Capelli and
Israël, 2007; Israël et al., 2004; Vercruyssen et al., 1989). Participant
whole-body motion can be further subdivided into active or passive
according to whether the movements are or are not intentionally
produced by the participant. Vercruyssen et al. (1989) studied the
effect of active motion (whole-body motion without displacement
in space) on time perception. Participants pressed a button to pro-
duce 10-s intervals while riding an ergometer cycle at a constant
velocity. The intervals produced while riding the cycle were shorter
than the intervals produced either before or after cycling, a result
that suggests that the internal clock ran faster while riding the
cycle.
Israël et al.’s (2004) study addressed the effects of passive
motion. Participants were blindfolded and then either moved pas-
sively forward and backward (pushed by a machine) or did not
move. In both conditions they were instructed to press a but-
ton once every second. Results showed that the production of 1-s
intervals was more variable in the condition with motion than
in the condition without motion. Furthermore, the frequency of
button presses was correlated with acceleration. Positive accelera-
tion reduced the inter response times (IRTs), negative acceleration
increased the IRTs, and no acceleration (no motion or motion at a
constant velocity) did not affect time production (see also, Capelli
et al., 2007). Similar effects were found during rotational motion
51
(e.g., Binetti et al., 2010) and after rotational motion (Capelli and
Israël, 2007).
Israël et al. (2004) suggested that time perception was disrupted
because their procedure was effectively a dual-task procedure
(Zakay, 1993), that is, a procedure in which two tasks were per-
formed simultaneously, one task being the perception of physical
displacement and the other task being the production of 1-s inter-
vals. The effect of motion on time perception could be due to the
interference of the non-temporal task on the concurrently per-
formed temporal task (e.g., Binetti et al., 2010; Capelli et al., 2007;
Capelli and Israël, 2007; Israël et al., 2004; Vercruyssen et al., 1989).
Other studies have shown that timing in dual-task conditions
tends to be more inaccurate than in single-task conditions. Thus,
produced or reproduced judgments differ more from the standard
and vary more across trials in dual than single tasks and, simi-
larly, the classification of standard durations is more error prone
and varies more across trials in dual than single tasks (e.g., Brown,
1997, 2006; Wearden et al., 2010). To illustrate, Hwang et al.
(2010) compared the performance of children and adolescents with
attention-deficit/hyperactivity disorder (ADHD) to that of a control
group without ADHD. The participants performed a reproduction
task under a dual-task paradigm in which the non-temporal task
involved counting visual stimuli presented on the computer screen.
The results showed that the judgments of ADHD children and ado-
lescents were less precise and accurate than those of the control
group. Moreover, the simultaneous performance of a non-temporal
task increased the number of time estimation errors in both
groups, although the errors were more pronounced in the ADHD
group.
Dual-task effects on timing occur also with animals. Lejeune
et al. (1999) exposed pigeons to two tasks, a temporal task involving
stimuli of different durations, and a non-temporal task involving
a variable-ratio (VR) reinforcement schedule. In the single-task
condition the pigeons performed only the temporal task and in
the dual-task condition they performed both tasks concurrently.
Specifically, in the temporal task the pigeons learned to choose one
of two keys following a Short stimulus, and to choose the other key
following a Long stimulus (bisection task). In the test phase, stimuli
with intermediate durations were also presented and the pigeons’
choices (‘Short’ or ‘Long’) were recorded. In the dual task condition,
the pigeons needed to respond during the stimulus presentation.
The percentage of ‘Long’ responses was lower when the pigeons
performed the two tasks concurrently than when they performed
only the temporal task.
To explain the results obtained with the dual-task paradigm,
Zakay (1993) proposed the Attentional Allocation Model. This model
states that attentional capacity is limited. Therefore, in a dual-
task condition, the non-temporal and temporal tasks compete for
these limited attentional resources such that, in the dual task,
the participant may “miss” parts of the to-be-estimated dura-
tion (see Lejeune et al., 1999). That is, with the split of resources
between the two tasks, attention to the standard may start only
after its onset, be suspended during its occurrence, or stop before
its offset. More generally, if accurate time judgments depend on
directed attention, then any competing task that takes attention
away from the temporal task will result in seemingly shorter stan-
dards (see also Zakay and Block, 1997). Phrased in terms of SET, the
Attentional Allocation Model states that in the dual-task condition
the pulses emitted during the standard may be lost and, conse-
quently, the standard may seem shorter than in the single-task
condition.
Casual observations suggest that humans may often engage in
dual tasks one of which involves a temporal judgment and the other
involves some form of motion. For example, a pedestrian may need
to estimate the time to cross a busy street safely while walking
or running. Given the high adaptive value of temporally regulated
52 A. Kroger-Costa et al. / Behavioural Processes 95 (2013) 50–59
behavior and given that humans are active agents often moving
while executing many tasks, the question arises, how does active
motion affect temporal discrimination? This then is the main ques-
tion asked in the present study.
To answer the question, we exposed subjects to a temporal
dual-task procedure (Israël et al., 2004). In the non-temporal task,
participants ran on a treadmill, and in the temporal task, they
performed either a temporal bisection task (Experiment 1), or a
temporal generalization task (Experiment 2), two common tasks
used to study timing in humans (e.g., Allan and Gibbon, 1991;
Wearden, 1991a). We compared the performance in the temporal
tasks when the participants were running on the treadmill (dual-
task condition) with performance when they were standing still
(single-task condition).
The literature on temporal dual-task procedures (e.g., Brown,
1997, 2008; Hwang et al., 2010; Zakay, 1993) indicates that the
non-temporal task disrupts the performance on the temporal task
and the Attentional Allocation Model states that this disruption may
be due to a loss of pulses during the standard in the dual-task
condition. Because a loss of pulses is equivalent to a decrease in
pacemaker speed, one could say that in the dual-task condition the
pacemaker was slower than in the single-task condition and for
that reason the subject judged the same standard as shorter in the
dual-task than in the single-task condition. However, the studies in
which the non-temporal task involves active, whole-body motion
show that produced intervals are generally shorter with motion
than without motion, a result more consistent with an increase in
pacemaker speed. For when the pacemaker emits pulses faster, any
reference number of pulses will be reached earlier, and for that rea-
son produced intervals will be shorter in the dual-task than in the
single-task condition.
The present study extended the analysis of the effects of active
motion to temporal discrimination tasks. If motion disrupts atten-
tion during the standard, with concomitant loss of pulses, then the
subject will judge the standard shorter in the dual-task than in the
single-task condition. In contrast, if motion increases the speed of
the pacemaker, the subject will judge the standard longer in the
dual-task than in the single-task condition.
2. Experiment 1
Experiment 1 compared time perception under the two con-
ditions of a dual-task procedure. In the single-task condition the
participants performed only a temporal bisection task, and in the
dual-task condition they performed the temporal bisection task
while running on a treadmill.
Walking or running on a treadmill removes the translational
component of motion, but preserves all the rotational-pendular
components. In particular, it preserves the fundamental biome-
chanical, perceptual and egomotion properties of regular motion.
In terms of biomechanical properties, free walking and walk-
ing on a treadmill are similar. In terms of perception, almost all
experiments on biological motion perception performed since the
seminal studies of Gunnar Johansson in the 1970s have removed
the translational component; nevertheless, recognition of motion
patterns and even complex actions remains straightforward for
both humans and animals. And in terms of egomotion percep-
tion, the vection experience (visual expansion) and the feeling
of acceleration may be reduced while walking on a treadmill,
but all the other multisensory motion cues including visual and
vestibular inputs coupled with rhythmic postural changes, audi-
tory (step sounds) and somatosensory (motion of the limbs) cues
remain intact (see Johansson, 1973, 1976). Hence, walking and
running on a treadmill should still be regarded as a locomotion
pattern.
Table 1
Order of each condition in each session for the two groups of participants. S = single
and D = dual task.
Order of conditions presentation
Session 1 Session 2
Group Condition 1 Condition 2 Condition 1 Condition 2
S-D:D-S Single-task Dual-task Dual-task Single-task
D-S:S-D Dual-task Single-task Single-task Dual-task
2.1. Method
2.1.1. Participants
Ten undergraduate students (5 females, mean age = 25.3,
SD = 2.0) from the University of Minho voluntarily participated in
the experiment. The participants had no previous experience with
the procedure and were blind to the hypothesis of the study.
2.1.2. Setting, apparatus, and stimuli
The experimental sessions were conducted in a 10 m × 7 m room
at the University of Minho (http://webs.psi.uminho.pt/lvp). The
room was equipped with a treadmill (Valiant 932900SE) and a
safety harness attached to the ceiling and used to prevent injuries
in case the participant fell off the treadmill (which never occurred
in the present studies). Approximately 20 cm in front of the tread-
mill was a table with a laptop computer, a pair of headphones,
and a wireless mouse. The computer presented the instructions,
controlled the stimulus presentations, and registered the mouse
responses.
2.1.3. Procedure
Before each session, the participants were asked to put on the
safety harness and the headphones, hold the wireless mouse in
their preferred hand, and stand up on the treadmill.
As Table 1 shows, each experiment comprised two sessions,
each approximately 25-min long, with a 2-h break between ses-
sions. During each session, there were two conditions, single-task
(S) and dual-task (D), each lasting approximately 10 min, with a
2-min break between them. In the single-task condition, the partic-
ipants performed only the temporal task. In the dual-task condition
the participants performed the temporal task simultaneously with
the non-temporal task. The difference between the two sessions
was the order of the conditions. The participants were divided into
two groups according to the order of the conditions: Group S-D:D-S
started with the single-task condition, and Group D-S:S-D started
with the dual-task condition.
At the beginning of the first session, the following instructions
were presented on the computer screen:
“Thank you for participating in our study. We are interested in
some features of behavior that are common to all people. More
specifically, we are interested in the sense of time that each
person has. I will present to you two tones, one short and one
long. Pay attention to both and tell me if you can hear them
clearly. When ready to begin, please say so.”
Next, the two anchor stimuli, a 300-ms (Short) and a 700-ms
(Long) tone, both 500-Hz in frequency, were presented in alterna-
tion, three times each, always starting with the short stimulus. After
the presentation of the anchor stimuli, the following instructions
were provided at the beginning of each condition:
“You will see a black square on the computer screen. Then you
will hear a tone. When the tone ends, the square will turn yellow.
At this moment, you will have to click one of the two mouse
buttons. If the tone seems LONG, click the RIGHT button of the
 A. Kroger-Costa et al. / Behavioural Processes 95 (2013) 50–59 53

Fig. 1. Individual and average proportion of “Long” responses plotted as a function of stimulus duration. The left and right panels show the data from groups S-D:D-S and
D-S:SD, respectively. The last row shows the average data. Open and close circles show the data from the single- and dual-task conditions, respectively. The lines are the
best-fit cumulative Gaussian (parameters in Table 2).
54 A. Kroger-Costa et al. / Behavioural Processes 95 (2013) 50–59

mouse. If the tone seems SHORT, click the LEFT button of the Table 2
Best-fitting parameters from cumulative Gaussian curve ( = mean,  = standard
mouse.
deviation) and variance accounted for (ω2 ) in Experiment 1. Ptc = participant.
Click any mouse button to start.” Group Ptc   ω2
Regardless of condition, single- or dual-task, these instructions Single Dual Single Dual Single Dual
and the presentation of the anchor stimuli occurred with the
S-D:D-S 01 55.55 74.89 475.28 498.59 0.99 0.98
treadmill off. The aim of these instructions was to ensure the learn- 03 31.99 47.78 425.23 401.75 1.00 0.98
ing of the correct responses to the anchor stimuli. The mapping 05 84.28 54.43 515.71 436.37 0.97 0.99
between the left and right responses and the short and long stimuli 07 94.54 102.98 527.81 460.46 0.99 0.86
was reversed for half of the participants. Henceforth, the correct 09 59.91 41.71 446.37 403.40 1.00 0.98

responses following the Long and Short stimuli will be designated D-S:S-D 02 51.42 67.20 471.14 457.10 0.99 0.98
“Long” and “Short”, respectively. 04 56.44 48.83 457.48 423.31 1.00 0.99
06 57.54 90.89 441.37 454.61 0.99 0.98
After reading the instructions and hearing the anchor stimuli,
08 52.93 51.72 428.38 428.88 0.98 0.99
the participants in the dual-task condition were asked to run on 10 34.85 64.68 498.76 492.12 1.00 0.99
a treadmill at a speed of 2 meters per second (m/s). This velocity
Average 57.95 64.51 468.75 445.66 0.99 0.97
was achieved by increasing the speed of the treadmill gradually
for about one minute. After the participants were running at the
target speed of 2 m/s for one additional minute, the temporal task
began. In the single-task condition, after reading the instructions single-task and dual-task conditions, respectively. The lines are the
and hearing the anchor stimuli, the participants waited two min- best-fit cumulative Gaussian curves. In the bottom panel, the lines
utes before the temporal task began. Once the temporal task began show the curves obtained from averaging the individual best-fit
the only difference between the two conditions was that the par- Gaussian curves. Table 2 shows the estimated parameters and the
ticipants were running in one case, and standing still in the other variance accounted for (ω2 ).
case. For both groups and conditions, the proportion of “Long”
In the temporal task, the two anchor stimuli and seven new responses tended to increase monotonically with stimulus dura-
stimuli (intermediate stimuli) were presented across trials in ran- tions from about 0 to about 1, which reveals good temporal
dom order. The intermediate stimuli, also 500-Hz tones, had the discrimination between the anchor stimuli. A mixed 9 × 2 × 2
following durations: 350, 400, 450, 500, 550, 600, and 650 ms. ANOVA, with groups (2 levels) as between-subjects factor, and
After the presentation of each stimulus, the participants had to stimulus duration (9 levels) and experimental condition (2 lev-
click one of the mouse buttons. Thus, one trial comprised the stim- els) as within-subjects factors, showed a significant effect of
ulus presentation and the participant’s response. None feedback duration, F(8, 64) = 218.93, p < .001, 2p = .96, and a significant inter-
was provided. The inter-trial interval (ITI) could be 1-s or 3-s long, action between stimulus duration and condition, F(8, 64) = 2.98,
randomly determined. p = .007, 2p = .27. All other effects were not significant. These
Each condition ended after 16 presentations of each anchor results revealed that the participants’ responses varied with stimu-
stimulus and 10 presentations of each intermediate stimulus, for lus duration and that the variation differed between experimental
a total of 102 trials. conditions.
To further examine how time perception varied with exper-
2.1.4. Data analysis imental condition, we compared the estimated parameters of
For each stimulus duration, t, we computed the correspond- the psychometric functions from the single-task and dual-task
ing proportion of “Long” responses, P(“Long”|t). These proportions conditions. The standard deviations ranged from 31.99 to 94.54
define the psychometric function, which typically follows a sig- (M = 57.95, SD = 19.20) in the single-task condition and from 41.71
moid curve, starting close to 0 at the shortest duration and ending to 102.98 (M = 64.51, SD = 20.00) in the dual-task condition (see
close to 1 at the longest duration. The curve is generally well fit by Table 2). A mixed 2 × 2 ANOVA comparing the standard deviations
a cumulative Gaussian function with two parameters, the mean, ( values) of the two groups as between-subjects factor and the two
, and the standard deviation, . The mean corresponds to the conditions as within-subjects factor, revealed no significant effect
Point of Subjective Equality (PSE), that is, the duration that yields of either factor or of their interaction (all Fs < 1.4, p > .27). Thus we
indifference between the “Long” and “Short” responses; hence, can reject the hypothesis that sensitivity to time was affected dif-
P(“Long”|t = ) = .5. The standard deviation is inversely related to ferentially by the experimental condition (single- or dual-task) or
the participants’ sensitivity to stimulus duration (smaller values of by the order in which the conditions were presented. Any interfer-
 correspond to steeper curves and higher temporal sensitivity). ence effects of running were not large enough to disrupt temporal
We used ANOVAs to compare the obtained psychometric discrimination.
functions and their estimated parameters across conditions and However, visual inspection of the average data suggests that the
sessions. Because a preliminary analysis showed no systematic curves of the two conditions had different PSEs. Individual PSEs (
differences between the two psychometric functions from the values) ranged from 425.23 to 527.81 ms (M = 468.75; SD = 35.81) in
single-task conditions and between the two psychometric func- the single-task condition and from 401.75 to 498.59 ms (M = 445.66,
tions from the dual-task conditions (see order in Table 1), we SD = 33.29) in the dual-task condition (see Table 2). A mixed 2 × 2
averaged them. Thus, each participant contributed two psychomet- ANOVA comparing the PSEs of the two groups (between factor) and
ric functions, one from condition Single and one from condition two conditions (within factor) revealed that the mean PSE in the
Dual. Significance level was set at p < .05. single-task condition were significantly different from the mean
PSE in the dual-task condition, F(1, 8) = 5.46, p = .048, 2p = .41, but
2.2. Results and discussion there was no effect of group or interaction between the two factors
(all Fs < 2.3, p > .17). This result indicates that, in general, under the
Fig. 1 shows the obtained psychometric functions. The left dual-task condition the participants had more “Long” responses
and right panels show the data for the S-D:D-S and D-S:S-D then in the single-task condition.
groups, respectively. The bottom graph shows the data averaged In sum, the results revealed that the participants’ responses
across groups. The open and closed circles are the data from the were not differentially sensitive to the stimulus durations in the
 A. Kroger-Costa et al. / Behavioural Processes 95 (2013) 50–59
single- and dual-task conditions, so concurrent execution of a non-
temporal task did not alter temporal sensitivity. However, as the
difference in the PSE between conditions and the significant inter-
action between stimulus durations and conditions revealed, the
participants perceived stimulus durations, particularly the inter-
mediate durations, as longer when they were running.
This result is consistent with Capelli et al. (2007), Israël et al.
(2004), and Vercruyssen et al. (1989) studies which found that,
when producing 1-s or 10-s intervals by pushing a button, passive
and active motion shortened the IRTs. Participants seem to perceive
the same standard as longer when moving than when remaining
stationary. However, there seems to be a difference between pas-
sive and active motion. With passive motion the effect did not occur
when there was no acceleration (i.e., no motion or motion at con-
stant velocity; e.g., Capelli et al., 2007; Israël et al., 2004), but with
active motion the effect was found under constant velocity (present
Experiment 1; Vercruyssen et al., 1989).
The magnitude of the shift in the psychometric function
obtained in the present study is within the range of the shifts
obtained with human participants when other variables are manip-
ulated in the temporal bisection task. One way to quantify
the magnitude of the shift is to divide the absolute difference
between the PSEs obtained in each condition by the average
PSE,
 
PSE single − PSE dual
× 100.
(PSE single + PSE dual)/2
In the present study, the magnitude of the shift was 5.05%. Gil
and Droit-Volet (2011) obtained a shift of 6.8% when investigat-
ing the effect of angry faces on time perception. Droit-Volet and
Wearden (2002) obtained a shift of 8.7% with 8-year olds when
investigating the effects of a flicker training on time perception.
And McCormack et al. (1999) obtained a shift of 6.9% when compar-
ing time perception in 5-year olds and undergraduate students. We
conclude that the effect obtained in the present study is consistent
with the effects obtained in other studies.
The present study cannot rule out an alternative account of the
overestimation effect. Instead of an increase in pacemaker speed,
the overestimation effect could be due simply to a shift of context,
form standing still while hearing the standard tones, to running in
the treadmill while hearing the test tones. To test this account, the
experimenter could present the standard tones while the partici-
pants were running and then present the test tones while they were
standing still. Effects in the opposite direction to that reported in the
present paper would be strong evidence for the pacemaker-based
account advanced above.
3. Experiment 2
The aim of Experiment 2 was to investigate the effect of motion
on temporal generalization. Participants were asked to evaluate if
a set of comparison tones varying in duration were equal to, or
different from, a standard tone. As in Experiment 1, we compared
performance on the generalization task under two conditions, a
dual-task condition in which the participants ran on a treadmill
while hearing and classifying the stimuli, and a single-task con-
dition in which the participants stood still on the treadmill while
hearing and classifying the stimuli.
For the single-task condition we expected a roughly bell-
shaped generalization gradient centered on the standard duration
(Wearden, 1991b). In the dual-task condition, if motion increases
subjective duration, as Experiment 1 suggested, then we expected
a gradient shifted to the left. For if active motion speeds the pace-
maker and thereby inflates durations, as it were, then durations
somewhat shorter than the standard should be perceived as equal
55
to the standard, but durations somewhat longer than the standard
should be perceived as even longer and therefore more unequal to
the standard than before; the net effect should be a leftward shift
in the generalization gradient.
3.1. Method
3.1.1. Participants
Ten volunteer undergraduate students from the University of
Minho participated in the experiment (six females and four males,
mean age = 25.2, SD = 1.54). The participants had no previous expe-
rience with the experimental procedure and were blind to the
hypothesis of the study.
3.1.2. Procedure
Similar to Experiment 1 (see Table 1) the participants were ran-
domly divided into two groups, one exposed to the single- and
dual-task conditions in the order S-D:D-S, and one exposed to the
two conditions in the order D-S:S-D. The two conditions, about
10 min each, were separated by a 2-min break. The two sessions,
each approximately 25-min long, were separated by a 2-h break.
At the beginning of the first session, the following instructions
were presented on the computer screen:
“Thank you for participating in our study. We are interested in
some features of behavior that are common to all people. More
specifically, we are interested in the sense of time that each
person has. I will present to you a CORRECT tone. Pay attention
to its duration and tell me if you can hear it clearly. When ready
to begin, please say so.”
After the instructions, the standard stimulus, a 500-Hz tone
lasting 500 ms, was presented five times. Next, a second set of
instructions was provided:
“You will see a black square on the computer screen. You will
hear a tone. When the tone ends, the square will turn yellow.
At this moment, you will have to click one of the two mouse
buttons. If the tone seems EQUAL to the CORRECT tone, click
the RIGHT mouse button. If the tone seems DIFFERENT, click the
LEFT button. Click any mouse button to start.”
The mapping between the right/left buttons and the
equal/different judgments was counterbalanced across partic-
ipants. Hence, instead of “Left” and “Right”, we refer to the two
responses as “Equal” and “Different”.
The instructions and the five presentations of the standard tone
were repeated at the beginning of each condition with the par-
ticipants standing still on the treadmill. Then, in the dual-task
conditions the treadmill was turned on in the manner described
in Experiment 1. In both conditions, the temporal generalization
task began 2 min later.
There were nine 500-Hz comparison stimuli varying in duration,
four shorter, one equal, and four longer than the 500-ms standard:
300, 350, 400, 450, 500, 550, 600, 650, and 700 ms. Henceforth,
the comparison stimulus with the same duration as the standard
will be named “equal-comparison”. After the presentation of each
stimulus, the participant had to press one of the mouse buttons. No
feedback was provided. As in Experiment 1, the stimulus presen-
tation and the response constituted a trial and the ITI equaled 1 or
3 s, randomly determined.
Each session ended after 128 trials, 64 in which the 500-
ms, equal-comparison duration was presented, and 64 in which
the 8 different-comparison durations were presented (for 8 trials
each). The order of stimulus presentations was randomized across
trials.
56 A. Kroger-Costa et al. / Behavioural Processes 95 (2013) 50–59

Table 3
Best-fitting parameters from cumulative Gaussian curve ( = mean,  = standard deviation) and variance accounted for (ω2 ) in Experiment 2. Ptc = participant.

Group Ptc  value  value A value ω2

Single Dual Single Dual Single Dual Single Dual

S-D:D-S 01 138.13 153.19 575.88 595.01 307.85 258.75 0.85 0.75

03 119.25 143.25 557.31 579.79 281.97 354.96 0.97 0.94
05 60.17 80.95 506.77 490.15 133.79 178.16 0.94 0.94
07 103.82 123.19 496.18 554.35 273.79 313.87 0.97 0.86
09 147.86 191.60 564.08 522.77 287.02 349.61 0.95 0.82

D-S:S-D 02 91.02 123.39 551.78 567.18 247.37 301.70 0.93 0.95

04 81.00 121.96 512.23 511.61 182.44 211.94 0.94 0.76
06 89.99 182.00 429.20 549.91 178.62 346.62 0.89 0.76
08 149.27 202.22 527.64 433.23 397.51 334.04 0.83 0.82
10 130.25 180.39 534.91 605.14 267.18 263.98 0.82 0.81

Average 111.08 150.21 525.60 540.91 255.75 291.36 0.91 0.84

3.1.3. Data analysis
The generalization gradients relating the proportion of “Equal”
responses to the stimulus duration were fitted by a Gaussian den-
sity function with equation
2
e−(1/2)((t−)/)
P(“Equal”|t) = A √ group (between) and condition (within) factors revealed a signif-
2
where P(“Equal”|t) is the proportion of “Equal” responses given a t-s
stimulus,  and  are the mean and standard deviation of the Gauss-
ian function, respectively, and A is a (vertically) scaling parameter
that, combined with , determines the overall probability of repor-
ting “Equal”. As in Experiment 1, a preliminary analysis showed no
significant differences between the two gradients from the single-
task conditions and between the two gradients from the dual-task
conditions (see order in Table 1), hence we averaged them. Thus,
each participant contributed two psychometric functions, one from
condition single and one from condition dual.
3.2. Results and discussion
Fig. 2 shows the individual and average proportion of “Equal”
responses plotted as a function of stimulus duration. The left panel
shows the data from the S-D:D-S group, and the right panel presents
the data from the D-S:S-D group. The open and closed circles show
the average data from the single- and dual-task conditions, respec-
tively. The lines show the best-fitting curves. In the bottom panel,
the lines show the averages of the individual curves. Table 3 shows
the corresponding parameters as well as the variance accounted
for.
To compare the generalization gradients we ran a mixed
9 × 2 × 2 ANOVA having the nine comparison durations and the
two conditions as within-subjects factors and the two groups as
between-subjects factor. The analysis revealed significant effects
of duration, F(8, 64) = 29.50, p < .001, 2p = .79, and interaction
between duration and condition, F(8, 64) = 2.58, p = .017, 2p = .24;
no other effects were significant (all Fs < 1, p > .5).
The proportion of “Equal” responses yielded generalization gra-
dients that were reasonably well fit by the Gaussian functions,
with ω2 values ranging from .75 to .97 (M = .87, SD = .08; see
Table 3). The  parameters ranged from 429.20 to 575.88 ms
(M = 525.6, SD = 42.79) in the single-task condition and from 433.23
to 605.14 ms (M = 540.91, SD = 52.57) in the dual-task condition.
A mixed 2 × 2 ANOVA comparing the individual  values of the
two groups (between factor) and the two conditions (within factor)
revealed no significant effect of condition, group or their interac-
tion (all Fs < .9, p > .3). Thus, contrary to the hypothesis, there were
no systematic differences in the mean of the gradients between the
single and dual-task conditions.
Concerning the width of the generalization gradients, the aver-
age data suggests that the curves from the dual-task condition
were flatter than the curves from the single-task condition. The
 parameters ranged from 60.17 to 149.27 (M = 111.08, SD = 30.5)
in the single-task condition and from 80.95 to 202.22 (M = 150.21,
SD = 38.65) in the dual-task condition. A mixed 2x2 ANOVA with
icant effect of condition, F(1, 8) = 47.16, p < .001, 2p = .85, because
the  values were smaller in the single-task condition (M = 111.08)
than the dual-task condition (M = 150.21), and a significant effect of
the interaction between group and condition, F(1, 1) = 6.51, p = .034,
2p = .45, because the difference mentioned above was bigger for
group D-S:S-D than group S-D:D-S; there was no effect of group
(F < .2, p > .6). These results suggest better temporal control under
the single-task condition. That is to say, running affected the pre-
cision with which participants identify the duration of intervals.
The individual A values ranged from 133.79 to 397.51
(M = 255.75, SD = 75.42) in the single-task condition and from
178.16 to 354.96 (M = 291.36, SD = 61.34) in the dual-task condition
(see Table 3). A mixed 2 × 2 ANOVA yielded no effect of condition,
group or interaction between these factors (all Fs < 2.7, p > .09).
We also fit the generalization gradients with a more specific
instantiation of SET (Church and Gibbon, 1982). Assume that the
subject’s representation of the standard is a Gaussian random vari-
able XS with mean  and standard deviation proportional to the
mean,  × . Moreover, when presented with a stimulus of duration
t, the participant responds “Equal” provided the relative difference
|(XS − t)/XS | is less than a threshold, b. Church and Gibbon (1982)
assumed a variable threshold, but here we explore the simpler case
of a constant threshold. From the previous assumptions, it follows
that the probability of an “Equal” response after a t-s stimulus is
P(“Equal”|t) = ˚(z2 ) − ˚(z1 )
where ˚(z) is the standard cumulative normal distribution, and z1
and z2 are defined as follows
1
 1 t
 1
 1 t

z1 = 1− , z2 = 1−
 1−b   1+b 
This specific model has three parameters (, , and b), the same
number as the Gaussian density model with a scale factor (, , and
A). In fact, the first two parameters of each model are equivalent,
for in both cases they characterize the mean and variability of the
representation of the standard. The third parameter, the threshold,
is similar to the scaling factor A, for increasing b or A increases
P(“Equal”|t). Given these similarities we expect the new model
to fit the data equally well, and to yield parameter values that do
not alter significantly the interpretations based on the Gaussian
density model. However, because the new model assumes a ratio
decision rule it can generate gradients that are not Gaussian
 A. Kroger-Costa et al. / Behavioural Processes 95 (2013) 50–59 57

Fig. 2. Individual and average proportion of “Equal” responses plotted as a function of stimulus duration. The left and right panels show the data from groups S-D:D-S and
D-S:SD, respectively. The last row shows the average data. Open and close circles show the data from the single- and dual-task conditions, respectively. The lines are the
best-fit Gaussian curves (parameters in Table 3).
58 A. Kroger-Costa et al. / Behavioural Processes 95 (2013) 50–59
and, in particular, gradients that are asymmetric (notice that the
difference between the two cumulative Gaussians, ˚(z2 ) − ˚(z1 ),
is not itself a Gaussian density).
The new model fit the data slightly better than the previ-
ous model: ω2 values averaged M = .90 (against M = .87). However,
statistical analyses (mixed ANOVAs) comparing the , , and b
parameters as a function of group (S-D:D-S vs. D-S:S-D) and condi-
tion (single task vs. dual task) yielded the exact same results as
before: neither the mean () nor the threshold (b) varied with
group or condition, but variability () increased in the dual task con-
dition, particularly for group D-S:S-D. In fact, the parameter values
of the two models were significantly correlated (range of correla-
tion coefficients, .68–.99). These results corroborate the idea that
running affected mainly the precision with which the participants
identified the duration of the intervals.
In sum, the generalization gradient from the dual-task condi-
tion was significantly wider than the gradient from the single-task
condition, which means that temporal control was reduced, partic-
ularly when the participants started the experiment by running on
the treadmill.
4. General discussion
The experiments reported above evaluated the effect of
active motion on time perception. In Experiment 1, participants
performed a temporal bisection task while running (dual-task con-
dition) or standing still (single-task condition). Results showed that
under the dual-task condition durations were perceived as longer
than under the single-task condition. In Experiment 2, participants
performed a temporal generalization task also while running or
standing still. Results showed that motion weakened temporal con-
trol. Taken together, these results indicate that the way motion
affects time perception depends on the temporal task.
In what follows, we compare our findings with the predictions
of the Attentional Allocation Model (e.g., Zakay, 1993; Zakay and
Block, 1997). Then, we describe, in more detailed, the compari-
son/decision component of SET and consider ways in which this
model may account for the effects of running. Finally we compare
the suggested account with the results from the temporal bisection
and generalization tasks.
4.1. Attentional Allocation Model
As mentioned before, according to Zakay (1993), in a dual-
task paradigm the non-temporal and temporal tasks compete for
limited attentional resources. The resulting split in the allocation
of attention reduces the storage of time clues, which in turn yields
a poorer representation of a temporal interval. Because temporal
judgments rely on the internal representations, it follows that, with
a reduced number of stored time cues, a shortening in the time
estimate occurs, and this in turn yields the perception of shorter
intervals in temporal discrimination tasks (Brown, 1997, 2008;
Zakay and Block, 1997). The model is consistent with the well-
documented effect that participants underestimate physical time
when performing a non-temporal task concurrently (e.g., Brown,
1997; Hemmes et al., 2004; Kladopoulos et al., 2004).
However, the results found in studies that used a whole-body
motion as the non-temporal task (Binetti et al., 2010; Capelli et al.,
2007; Capelli and Israël, 2007; Israël et al., 2004; Vercruyssen et al.,
1989), as well as the results obtained in the present study, are the
opposite of the results predicted by the Attentional Allocation Model.
Under a concurrent non-temporal task that involved whole-body
motion (passive or active), the participants perceived time as longer
rather than shorter, that is, they overestimated physical time.
Our results agree with Molet et al. (2011) finding that the effect
of the non-temporal task varies with the nature of the task. Their
findings show that when the non-temporal task required counting
backwards by threes, physical time was underestimated. However,
when the non-temporal task involved exerting continuous force on
a transducer, the opposite effect was obtained, physical time was
overestimated.
4.2. Scalar Expectancy Theory (SET)
The effects found under active motion may be interpreted in the
light of Scalar Expectancy Theory (SET). In the temporal bisection
task, the pulses accumulated during the short and long samples are
stored in separate memories. In the temporal generalization task,
the pulses accumulated during the standard stimulus are stored in
the same memory. To decide whether a test stimulus is closer to
the short or to the long samples (temporal bisection), or equal to
or different from the standard sample (temporal generalization),
the participant compares the pulses accumulated during the test
stimulus with samples extracted from memory. The decision rule
is slightly different in the two tasks. Thus, in the bisection task,
SET assumes that the participant forms two ratios Xt /XS and XL /Xt ,
where Xt is the number of pulses in the accumulator at the end of
the test stimulus, XS is a sample extracted from the memory asso-
ciated with the short sample, and XL is a sample extracted from
the memory associated with the long sample. If (Xt /XS ) < ˇ(XL /Xt ),
where ˇ is a bias parameter, the participant responds “Short”; oth-
erwise it responds “Long”. In the temporal generalization task, the
participant forms the ratio Xt /XS , where Xt is the number of pulses
in the accumulator at the end of the test stimulus and XS is a sample
extracted from the memory associated with the standard duration.
It responds “Equal” provided the ratio |XS − Xt |/XS < ˇ, and “Dif-
ferent” otherwise (e.g., Bangert et al., 2011; Gibbon, 1977, 1991;
Grondin, 2010; Lejeune and Wearden, 2006; Wearden, 1991b).
If we assume that, in a temporal dual-task paradigm (when
cognitive tasks of the type adding or counting numbers are imple-
mented as the non-temporal task), divided attention causes loss
of pulses in the accumulator, then the result is a shortening of
perceived stimulus durations. But if we assume that non-temporal
task such as running accelerate the pacemaker, then more pulses
will be accumulated and the result will be a lengthening of
perceived duration. Perhaps, then, the nature of the concurrent
task will determine which effect will prevail. Tasks with or without
motion may have different effects on the pacemaker–accumulator
unit (loss of pulses or increase in rate of pulse generation).
We consider now how the foregoing ideas might account for
the empirical findings. In Experiment 1 the participants perceived
stimulus durations as longer when they performed the two
tasks concurrently (temporal discrimination + running) compared
to when they performed only the temporal task. This result can be
explained by assuming an increase in the speed of the pacemaker
– the same physical duration will yield a higher value of Xt and
therefore a higher probability of classifying the stimulus as “Long”.
However, in a temporal generalization task, an increase in pace-
maker speed will shift the baseline generalization gradient to the
left. This prediction was not observed in Experiment 2, which
showed a flattening rather than a shift of the generalization gra-
dient during running. Before advancing an explanation for the
flattening of the generalization gradient, consider what SET pre-
dicts if the pacemaker speed increases but the participant updates
its memory for the standard duration with the new pacemaker
speed. That is, suppose that the memory for the standard duration
continues to be updated while the participant is running. In this
case, the participant in the running condition is similar to the par-
ticipant in the standing still condition in all respects except one, its
pacemaker is faster. But in this case SET predicts no effect of running
on the generalization gradient. In fact, it can be shown (Gibbon and
Church, 1984) that according to SET, the temporal generalization
 A. Kroger-Costa et al. / Behavioural Processes 95 (2013) 50–59 59

gradient does not depend on the absolute value of the pacemaker Capelli, A., Deborne, R., Israël, I., 2007. Temporal intervals production during passive
speed. self-motion in darkness. Curr. Psychol. Lett.: Behav. Brain Cogn. 22, Retrieved
from http://cpl.revues.org/index2672.html
There are at least two ways to predict a flattening of the gradient Capelli, A., Israël, I., 2007. One second interval production task during post-rotatory
with running. The most straightforward is to assume that the vari- sensation. J. Vestib. Res. 17, 239–249.
ability of the internal clock, represented in SET by the coefficient Church, R.M., Gibbon, J., 1982. Temporal generalization. J. Exp. Psychol. Anim. Behav.
Process. 8, 165–186.
of variation of the pacemaker speed, increases with running. The Droit-Volet, S., Wearden, J.H., 2002. Speeding up an internal clock in children?
problem with this hypothesis, however, is that it predicts a decrease Effects of visual flicker on subjective duration. Q. J. Exp. Psychol. 55, 193–211,
in the slope of the psychometric function in the bisection task when http://dx.doi.org/10.1080/02724990143000252.
Gibbon, J., 1977. Scalar expectance theory and Weber’s law in animal timing. Psychol.
the participant is running, a decrease that was not observed. Rev. 84, 279–325, http://dx.doi.org/10.1037/0033-295X.84.3.279.
Another way to predict a flattening of the gradient is to assume Gibbon, J., 1991. Origins of scalar timing theory. Learn. Motiv. 22, 3–38,
that, during testing, the participant’s memory for the standard http://dx.doi.org/10.1016/0023-9690(91)90015-Z.
Gibbon, J., Church, R.M., 1984. Sources of variance in an information processing the-
duration is contaminated by other test samples, both shorter and
ory of timing. In: Roitblat, H.L., Bever, T.G., Terrace, H.S. (Eds.), Animal Cognition.
longer than the standard. In this case, the denominator of the deci- Erlbaum, Hillsdale, NJ, pp. 465–487.
sion ratio would be equal to XS on some trials, greater than XS on Grondin, S., 2010. Timing and time perception: a review of recent behavioral and
other trials (the effect of test samples longer than the standard), and neuroscience findings and theoretical directions. Atten. Percept. Psychophys. 72,
561–582, http://dx.doi.org/10.3758/APP.72.3.561.
smaller than XS on still other trials (the effect of test samples shorter Hemmes, N.S., Brown, B.L., Kladopoulos, C.N., 2004. Time perception with and with-
than the standard). The combined effect of the memory contami- out a concurrent nontemporal task. Atten. Percept. Psychophys. 66, 328–341,
nation will be a decrease in the proportion of “Equal” responses in http://dx.doi.org/10.3758/BF03194882.
Hwang, S., Gau, S.S., Hsu, W., Wu, Y., 2010. Deficits in interval timing measured
the region of the standard duration and an increase in the propor- by the dual-task paradigm among children and adolescents with attention-
tion of “Equal” responses in the regions of the durations shorter deficit/hyperactivity disorder. J. Child Psychol. Psychiatry 51, 223–232,
and longer than the standard. In a word, the gradient will flatten http://dx.doi.org/10.1111/j.1469-7610.2009.02163.x.
Israël, I., Capelli, A., Sablé, D., Lauret, C., Lecoq, C., Bredin, J., 2004. Multifactorial inter-
by decreasing in the middle region and increasing in the tails, the actions involved in linear self-transport distance estimate: a place for time. Int.
observed effect. J. Psychophysiol. 53, 21–28, http://dx.doi.org/10.1016/j.ijpsycho.2004.01.002.
The foregoing hypothesis suggests that running may have task Johansson, G., 1973. Visual perception of biological motion and a model for its anal-
ysis. Percept. Psychophys. 14, 201–211, http://dx.doi.org/10.3758/BF03212378.
specific effects or, more precisely, that it may have multiple effects Johansson, G., 1976. Spatio-temporal differentiation and integration in visual motion
but not all of them would be expressed in a specific task. On the one perception. Psychol. Res. 38, 379–393.
hand, speeding the pacemaker would be expressed in the bisec- Kanai, R., Paffen, C.L.E., Hogendoorn, H., Verstraten, F.A.J., 2006. Time dilation in
dynamic visual display. J. Vis. 6, 1421–1430, http://dx.doi.org/10.1167/6.12.8.
tion task by the increase in the proportion of “Long” responses,
Kladopoulos, C.N., Hemmes, N., Brown, B.L., 2004. Prospective timing under
but it would not be expressed in the temporal generalization task dual-task paradigms: attentional and contextual-change mechanisms. Behav.
because the memory for the standard would also change when the Process. 67, 221–233, http://dx.doi.org/10.1016/j.beproc.2003.12.004.
speed of the pacemaker increases. On the other hand, memory con- Lejeune, H., Wearden, J.H., 2006. Review article. Scalar properties in ani-
mal timing: conformity and violations. Q. J. Exp. Psychol. 11, 1875–1908,
tamination would be expressed in the temporal generalization task http://dx.doi.org/10.1080/174702107012822576.
by the flattening of the gradient, but it would not be expressed in Lejeune, H., Macar, F., Zakay, D., 1999. Attention and timing: dual-
the bisection task because its effect is canceled when two memory task performance in pigeons. Behav. Process. 45, 141–157,
http://dx.doi.org/10.1016/S0376-6357(99)00015-7.
stores are involved. The foregoing hypothesis opens new avenues Matthews, W.J., 2011. Time winding down: how do changes in speed affected the
for research on the effects of running on time perception. perception of duration? J. Exp. Psychol. Hum. Percept. Perform. 37, 1617–1627.
McCormack, T., Brown, G.D.A., Maylor, E.A., Darby, R.J., Green, D., 1999. Developmen-
tal changes in time estimation: comparing childhood and old age. Dev. Psychol.
Acknowledgements 35, 1143–1155, http://dx.doi.org/10.1037/0012-1649.35.4.1143.
Michon, J.A., 1993. Concerning the time sense: the seven pillars
Kroger-Costa, A is supported by doctoral Grants from the Por- of time psychology. Psychol. Belg. 33, 329–345, Retrieved from
http://www.jamichon.nl/jam writings/1993 pillars.pdf
tuguese Foundation for Science and Technology (Fundação para a Molet, M., Alessandri, J., Zentall, T.R., 2011. Subjective time: cognitive and physi-
Ciência e Tecnologia, FCT) (SFRH/BD/66728/2009). cal secondary tasks affect timing differently. Q. J. Exp. Psychol. 64, 1344–1353,
http://dx.doi.org/10.1080/17470218.2011.552728.
Richelle, M., Lejeune, H., 1980. Time in Animal Behavior. Pergamon Press, Oxford,
References UK.
Vercruyssen, M., Hancock, P.A., Mihaly, T., 1989. Time estimation performance,
Allan, L.G., Gibbon, J., 1991. Human bisection at the geometric mean. Learn. Motiv. before, during, and following physical activity. J. Hum. Ergol. (Tokyo) 18,
22, 39–58, http://dx.doi.org/10.1016/0023-9690(91)90016-2. 167–179.
Bangert, A.S., Reuter-Lorenz, P.A., Seidler, R.D., 2011. Dissecting the clock: under- Wearden, J.H., 1991a. Human performance on an analogue of
standing the mechanisms of timing across tasks and temporal intervals. Acta an interval bisection task. Q. J. Exp. Psychol. 43, 59–81,
Psychol. (Amst.) 136 (1), 20–34, http://dx.doi.org/10.1016/j.actpsy.2010.09.006. http://dx.doi.org/10.1080/14640749108401259.
Bindra, D., Wakesberg, H., 1956. Methods and terminology in studies of time esti- Wearden, J.H., 1991b. Do humans possess an internal clock
mation. Psychol. Bull. 53, 155–159. with scalar timing properties? Learn. Motiv. 22, 59–83,
Binetti, N., Siegler, I.A., Bueti, D., Doricchi, F., 2010. Time in motion: effects of whole- http://dx.doi.org/10.1016/0023-9690(91)90017-3.
body rotatory accelerations on timekeeping processes. Neuropsychologia 48, Wearden, J.H., 2008. The perception of time: basic research and some
1842–1852, http://dx.doi.org/10.1016/j.neuropsychologia.2010.03.009. potential links to the study of language. Lang. Learn. 58, 149–171,
Brown, S.W., 1997. Attentional resources in timing: interference effects in concur- http://dx.doi.org/10.1111/j.1467-9922.2008.00468.x.
rent temporal and non-temporal working memory tasks. Percept. Psychophys. Wearden, J.H., O’Rourke, S.C., Matchwick, Z.M., Maeers, S., 2010. Task
59, 1118–1140. switching and subjective duration. Q. J. Exp. Psychol. 63, 531–543,
Brown, S.W., 2006. Timing and executive function: bidirectional interference http://dx.doi.org/10.1080/17470210903024768.
between concurrent temporal production and randomization tasks. Mem. Cogn. Zakay, D., 1993. Time estimation methods: do they influence prospective durations
34, 1464–1471. estimates? Perception 22, 91–101, http://dx.doi.org/10.1068/p220091.
Brown, S.W., 2008. The attenuation effect in timing: counteracting dual-task Zakay, D., Block, R.A., 1997. Temporal cognition. Curr. Dir. Psychol. Sci. 6, 12–16,
interference with time-judgment skill training. Perception 37, 712–724, http://dx.doi.org/10.1111/11467-8721.ep11512604.
http://dx.doi.org/10.1068/p5698.